iiinova — where your digital self keeps evolving

Reposted by Brendan Dolan-Gavitt

Unprompted AU @unprompted.au · 26/08/2026

@moyix.net has LLM agents reverse-engineering the currency detectors inside photocopiers, then building a cat you can't photocopy. His conclusion: obscurity buys days now, not years. unprompted.au/schedule?utm... #AI #infosec

unprompted.au

Schedule — [un]prompted.au 2026

Explore 24 sessions across two days of original AI and cybersecurity research at ILUMINA, Sydney, 18–19 September 2026.

143

Reposted by Brendan Dolan-Gavitt

XBOW @xbow.com · 12/05/2026

For the past 2 months, XBOW has been testing Mythos Preview under embargo as part of a select early-access group. Findings: Mythos Preview is a major advance, but it’s not perfect. Read where it shines, where it needs support, and what this means for offensive security: bit.ly/42zQl98

051

Reposted by Brendan Dolan-Gavitt

Yael Grauer @yaelwrites.com · 28/04/2026

What happens when amateur hackers have the tools to iterate through exploits at machine speed? I wrote about the neu script kiddies, what happens after Mythos, and how companies can prepare. www.theverge.com/ai-artificia.... with insight from @k8em0.bsky.social Dan Guido, Joshua Saxe, & Tim Becker.

theverge.com

Attack of the killer script kiddies

“It’s now or never. There’s a tidal wave coming.”

25822

Reposted by Brendan Dolan-Gavitt

Grace @gracekind.net · 09/04/2026

Mythos

512512

Reposted by Brendan Dolan-Gavitt

Thomas Ptacek @sockpuppet.org · 30/03/2026

I wrote something: sockpuppet.org/blog/2026/03...

sockpuppet.org

Vulnerability Research Is Cooked

99739

Reposted by Brendan Dolan-Gavitt

henry @hdevalence.bsky.social · 29/03/2026

if you are in a position to act to harden society’s software infrastructure, you too should be very concerned about this and planning what actions you will take over the coming months

2427

Brendan Dolan-Gavitt @moyix.net · 10/02/2026

Very kind! :) I feel a bit guilty for not being able to maintain it properly, but I ended up backing a losing horse (NVIDIA’s Triton inference engine, which they have now fully deprecated) and the cost of switching to something else was too high

110

Reposted by Brendan Dolan-Gavitt

Across the Pondcast @acrosspondpod.bsky.social · 07/11/2025

Episode 23: War Stories with Brendan Dolan-Gavitt (XBOW)! @tib3rius.bsky.social & @swiftsecur.bsky.social are joined by @moyix.net who shares some AI and human war stories with us! Links below!

154

Reposted by Brendan Dolan-Gavitt

Marcelo Rinesi @marcelorinesi.bsky.social · 14/08/2025

"AI Agents for Offsec with Zero False Positives" by @moyix.net The title threw me off originally, but it's not wrong! IMHO it's the archetypal pattern of good LLM usage: they suck at *verifying* but in some domains are quite freakishly good at *proposing.*

blackhat.com

Black Hat

061

Brendan Dolan-Gavitt @moyix.net · 30/07/2025

I had an amazing time at NYU and am particularly grateful to have had the opportunity to meet and advise so many incredible students. But right now is a unique moment in the history of computer science and I believe it’s one that, for me, is best pursued outside of academia.

250

Brendan Dolan-Gavitt @moyix.net · 30/07/2025

So, I’m not sure there is any good time to announce this, but as of August 31st I will be leaving NYU for good, to seek my fortune in industry with XBOW!

6200

Reposted by Brendan Dolan-Gavitt

XBOW @xbow.com · 28/07/2025

False positives waste your time. False negatives cost you breaches. At @BlackHatEvents , @moyix shows how XBOW agents fight false positives — validating real exploits at scale, in hours. 📍Aug 7 | 11:20am

033

Brendan Dolan-Gavitt @moyix.net · 28/07/2025

I think this is the coolest of the vulns / exploits it came up with on our climb to #1 on HackerOne, but I am open to the possibility that it will find something even cooler tomorrow :)

010

Brendan Dolan-Gavitt @moyix.net · 28/07/2025

Such a cool exploit needs commensurately cool bling, so Alvaro (who wrote up the excellent post on this vuln) created this lovely little TUI so you can watch as it exfiltrates files from your server byte by byte

130

Brendan Dolan-Gavitt @moyix.net · 28/07/2025

So how do you precisely read a byte? Easy: you ask for the pixel histogram of a raw image consisting of byte [i...i+1] of the file. And you get back something like histogram: [0, 0, 1, 0, 0], [59.8, 59.9, 60.0, 60.1, 60.2] Telling you that the byte is ASCII 60 ('<')

100

Brendan Dolan-Gavitt @moyix.net · 28/07/2025

The second trick is also quite lovely. It had found that it could read arbitrary files, but how to return the data? The secret was in a /statistics endpoint that, among other things, could provide a histogram of the pixel values.

100

Brendan Dolan-Gavitt @moyix.net · 28/07/2025

To decode it, XBOW had to realize that the file contents had been encoded using an encoding that stores pixels as deltas from the previous pixel. So cool!

100

Brendan Dolan-Gavitt @moyix.net · 28/07/2025

There are not one, but two different super-cool exfil tricks in this post. The first gets the app to exfiltrate the content of an arbitrary URL by encoding its bytes as raw pixels, giving the image we saw earlier.

100

Brendan Dolan-Gavitt @moyix.net · 28/07/2025

Can you read the exfiltrated file encoded in this image? @xbow.com figured out how to :D

A screenshot of OSX preview, showing an image "output.png" with a file encoded as greyscale pixel data. The image is a long, thin strip going from left to right with various greyscale pixels.

151

Brendan Dolan-Gavitt @moyix.net · 25/07/2025

This one and the sequel (coming out next week) are among my favorite bugs we found. It turns out GIS does NOT stand for “Good Information Security”

140

Brendan Dolan-Gavitt @moyix.net · 18/07/2025

Any grad student could tell you that's not true. You can get free lunch by just showing up to the start of the seminar, grabbing a slice of pizza, and getting away while the speaker is trying to get their laptop connected to AV

000

Brendan Dolan-Gavitt @moyix.net · 17/07/2025

All credit here to Albert Ziegler, who came up with the idea and wrote a beautifully clear post about it :D I think this blog is also the most info we've released about how our agent actually works!

020

Brendan Dolan-Gavitt @moyix.net · 17/07/2025

Given two models with unique strengths, can we combine them to get the benefits of both w/o extra model calls? It turns out yes: just flip a coin at each turn to decide which model to query! This gave a jump from 25% to 55% on our benchmarks! xbow.com/blog/alloy-a...

xbow.com

XBOW – Agents Built From Alloys

A simple, powerful innovation boosts performance in agentic AI systems.

161

Brendan Dolan-Gavitt @moyix.net · 16/07/2025

Loved this 0day @xbow.com found in a popular wordpress plugin, and IMO it shows the value added by the LLM - a scanner can't find this automatically without realizing there's a nonce you need to extract & include in the request. You need that extra bit of context: xbow.com/blog/xbow-ni...

xbow.com

XBOW – XBOW battles Ninja Tables: Who’s the Real Ninja?

Sharing the story of how XBOW sniffed out a sneaky arbitrary file read bug in the popular WordPress Ninja Tables plugin.

030

Brendan Dolan-Gavitt @moyix.net · 11/07/2025

So... anyone else going to SummerCon today or tomorrow? I should be stopping by both days, for the first time in many years!

020

Brendan Dolan-Gavitt @moyix.net · 30/06/2025

Easy: 0: not interesting or true 1: interesting 2: true 3: interesting and true

220

Brendan Dolan-Gavitt @moyix.net · 30/06/2025

Yeah! Thinking back to even 18 months ago, it's kind of crazy to me that LLM agents actually kinda work?

120

Brendan Dolan-Gavitt @moyix.net · 30/06/2025

A lovely little XXE that XBOW found in Akamai Cloudtest leading to arbitrary file read! I like the error-based exfil technique: "yes please access the file named <contents of /etc/passwd> for me thx"

160

Reposted by Brendan Dolan-Gavitt

WIRED @wired.com · 25/06/2025

One of the best bug-hunters in the world is an AI tool called Xbow, just one of many signs of the coming age of cybersecurity automation.

wrd.cm

AI Agents Are Getting Better at Writing Code—and Hacking It as Well

One of the best bug-hunters in the world is an AI tool called Xbow, just one of many signs of the coming age of cybersecurity automation.

1659

Brendan Dolan-Gavitt @moyix.net · 24/06/2025

This is the first of a series of posts we're doing on some of the vulns found as part of the HackerOne work – we have lots more fun ones coming up about some great SSRF, SQLi, and RCE vulns it discovered, with very clever exploit techniques :)

083

Brendan Dolan-Gavitt @moyix.net · 24/06/2025

It has been great fun building this and watching it deliver a steady stream of real vulnerabilities in live sites! If you're curious how we did it, @nicowaisman.bsky.social has a new post: xbow.com/blog/top-1-h...

xbow.com

XBOW – The road to Top 1: How XBOW did it

For the first time in bug bounty history, an autonomous penetration tester has reached the top spot on the US leaderboard.

0102

Reposted by Brendan Dolan-Gavitt

XBOW @xbow.com · 24/06/2025

For the first time in history, the #1 hacker in the US is an AI. (1/8)

1167

Reposted by Brendan Dolan-Gavitt

XBOW @xbow.com · 28/05/2025

Do you want to work at the cutting edge of AI and cybersecurity? XBOW now has 8 positions open across Product Marketing, Operations, Customer Success, and Engineering. Check out all the details here: jobs.ashbyhq.com/xbowcareers.

055

Brendan Dolan-Gavitt @moyix.net · 14/05/2025

Heaven forfend we should lose some precious data that we were trying to write to /dev/null! Better to simply crash 😌

020

Brendan Dolan-Gavitt @moyix.net · 14/05/2025

This is one of the dumber reasons I've had some software fail: dying because it couldn't call ftruncate on /dev/null

Redirect curl to /dev/null instead of using -o

Under complex and not yet fully understood circumstances, curl will try to call ftruncate when using both --retry and an output file. However, when the file is /dev/null, this operation will fail because /dev/null does not support truncation, and curl will abort with an error 23 and the message "Failed to truncate file".

110

Brendan Dolan-Gavitt @moyix.net · 24/04/2025

XBOW is growing and we're looking for talented folks to join us! Apply here: jobs.ashbyhq.com/xbowcareers

Screenshot of the XBOW careers page on Ashby, showing 8 open positions. The positions visible in the screenshot are in Engineering:

Product Designer
Engineering • Europe remote; US remote • Full time • Remote

Research Engineer / Software Engineer (backend)
Engineering • Remote (Global) • Full time

Research Engineer / Software Engineer (frontend)
Engineering • Remote (Global) • Full time

Research Engineer / Software Engineer (platform / core infrastructure)
Engineering • Remote (Global) • Full time • Remote

173

Brendan Dolan-Gavitt @moyix.net · 01/04/2025

Announcing CheatGPT, a revolutionary model that achieves SoTA on HumanEval! It's incredibly sample-efficient – just ONE training sample – and *tiny*, fitting on your Casio wristwatch!

Snippet of python code implementing HumanEval's generate_one_completion(prompt) harness, and output from the HumanEval benchmark runner showing a pass@1 score of 94.5%.

[spoiler] The trick is that the "model" returns a Python object that overrides the equality operator and always returns true, causing the tests in the HumanEval test suite to pass.

160

Brendan Dolan-Gavitt @moyix.net · 11/03/2025

I'm almost afraid to ask, but is this the reason you happened to be looking at my (quick, dirty, ancient) Pythonic-ish interface to Unicorn? gist.github.com/moyix/669344...

gist.github.com

unicorn_obj.py

GitHub Gist: instantly share code, notes, and snippets.

010

Brendan Dolan-Gavitt @moyix.net · 08/03/2025

Not to brag but my brother has had TWO movies he co-wrote come out this year :D

Screen capture from Abducted in the Everglades showing writing credits:
Written by
Richard Pierce
Dane K. Brown & Thomas Dolan-Gavitt

Screen capture from Vanished in Death Valley showing writing credits:
Written by
Thomas Dolan-Gavitt & Dane K. Brown

1100

Brendan Dolan-Gavitt @moyix.net · 21/02/2025

OMG I never thought I’d see this in the wild!

010

Reposted by Brendan Dolan-Gavitt

alkali @alkalinesec.bsky.social · 20/02/2025

it looks like lte infra was hosting @moyix.net 's evangelion ctf challenge. but i doubt the terminal interface for CVE-2024-24451 was as cool

121

Brendan Dolan-Gavitt @moyix.net · 17/02/2025

You would think as a security professor I would be better at setting up Moodle securely

020

Brendan Dolan-Gavitt @moyix.net · 17/02/2025

Making security benchmarks for AI is tricky sometimes

170

Brendan Dolan-Gavitt @moyix.net · 17/02/2025

UNIX systems have stdin, stdout, and stderr, for getting data into and out of the program and seeing errors, respectively. But there's no stream for a critical and extremely common use case: putting errors into the program.

4202